Neuro-Fuzzy-Based Adaptive Control for Autonomous Drone Flight

Adaptive control is the capability of a control system to modify its operation and achieve the best possible operation mode. A quadcopter is a nonlinear, unstable and under-actuated dynamic system, thus providing a challenge to control engineers in controlling and stabilising it during flight. This paper proposes the design, development, and application of an intelligent adaptive hybrid controller to control and stabilise the drone. The training data for adaptive neuro-fuzzy inference systems (ANFIS) are generated by the Linear Quadratic Regulator (LQR) under white-noise disturbance. The trained ANFIS is subsequently used to estimate the parameters of the control distribution matrix for the actual fault condition and the reconfiguration is carried out by computing new feedback gain using the pseudo-inverse technique. For the simple adaptive controller, LQR is also used to generate the desired trajectories of the reference model. In both experiments, the extended Kalman filter is implemented due to its non-linearity benefit. We demonstrate the performance of the proposed approach as a representative case study. The preliminary numerical simulation results further indicate that the proposed method is promising compared to conventional control techniques to control and stabilise a quadcopter drone.